An improved method for eliminating false matches
Zhang Songtao, Chao Li, Li Liqing · 2017
The Eliminating of false feature matches is an important follow-up work of feature detection extraction and matching, which is of great significance in improving the quality of image matching. In this paper, it experimentally analyzed several kinds of false feature matches rejection methods (Distance-ratio Criterion method, Bi-direction Matching method and RANSAC), which are performed based on the ORB feature point detection and descriptor extraction and the BF matching. In view of the overall filtering effect, RANSAC is relatively optimal, but it still has the disadvantages of slower filtering speed and fewer matching points after filtering. Propose an improved algorithm based on Bi-direction Matching method and Distance-ratio Criterion method, and conduct series of experiments with it. The experiments show that the filtering effect of the improved filtering algorithm is better than RANSAC. It is less time-consuming (reduced by 58.4%, compared with RANSAC), and retains more correct matches (increased by 14.8%, compared with RANSAC) while removing the false matches.